新功能
遇到模型限流聰明等待，額度耗盡立刻停手
若是頻繁發問而觸發了雲端模型的呼叫頻率限制，命令列工具現在會更聰明地處理重新嘗試的時間。以前只要遇到 API（程式與雲端模型溝通的管道）回傳需要稍候重試，工具一律死板地固定等待五秒。現在它會完全照著伺服器要求的秒數等待。如果伺服器要求等待超過三十秒，或是回報每日額度與帳單上限已滿，工具便會立刻停止執行，不再盲目重試。
繁體中文
效能優化
提升對話畫面繪製效能，捲動與運作時更省資源
在快速捲動落落長的對話紀錄、或看著代理人（代你執行任務的 AI）忙碌時，介面操作現在感覺更加輕快。對話畫面在重新繪製時的效能獲得了改善。以前每當畫面快速重繪，電腦的處理器（負責執行所有運算的核心硬體）使用率容易飆高，記憶體也會頻繁配置與釋放。現在無論是快速捲動視窗還是等待代理人即時輸出，都能有效減少運算資源耗損與記憶體負擔。
新功能
產物檢視器按鍵提示更明確，切換模式一目了然
想要預覽 AI 生成的圖表或公式等產物（獨立生成的圖表、公式或文件檔案）時，底部切換顯示模式的提示變得更直覺了。以前畫面上的 m 按鍵提示只寫 toggle ASCII 或 toggle raw，讓人難以預期切換後的畫面。現在提示文字會直接寫出內容種類與目標模式，例如清楚標示 diagram ASCII、diagram source 或是 math image。
問題修復
窄終端機視窗文字不再被截斷，長行與標籤自動排版
把終端機視窗縮小到約 40 欄寬的狹窄寬度時，各個功能面板的文字再也不會因為超出邊界而被裁切。以前在工作區信任對話框、/help 說明面板，以及登入與 MCP（讓 AI 連接外部工具的協定）驗證畫面中，長文字容易直接消失在螢幕邊緣。現在這些畫面中的長行文字與操作導引提示都會自動折行顯示，/help 的分頁標籤列也會自動緊縮排版以完整容納，排版截斷問題已經修好。
English
FEATURE
Smarter Retry Delays on Rate Limits, Immediate Stop When Quota Exhausted
If frequent prompts trigger cloud model rate limits, the command-line tool now handles retry delays more intelligently. Previously, whenever the API returned a retry-after response, the tool rigidly waited for a fixed five seconds. Now, it waits for the exact number of seconds requested by the server. If the server requests a delay of more than thirty seconds, or reports that the daily quota or billing limit is exhausted, the tool stops immediately instead of retrying blindly.
PERF
Improved Chat Screen Rendering Performance for Lower Resource Usage While Scrolling
When quickly scrolling through lengthy conversation histories or watching an agent work, UI interactions now feel significantly snappier. Redrawing performance for the conversation screen has been improved. Previously, rapid screen repaints often spiked CPU usage and caused frequent memory allocations and releases. Now, whether scrolling quickly through the window or waiting for real-time agent output, compute overhead and memory pressure are both substantially reduced.
FEATURE
Clearer Key Hints in the Artifact Viewer for Intuitive Mode Toggling
When previewing AI-generated artifacts such as standalone diagrams, formulas, or document files, the bottom hint for toggling display modes is now much more intuitive. Previously, the on-screen m key hint only displayed toggle ASCII or toggle raw, making it hard to anticipate what the screen would switch to. Now, the prompt text explicitly shows the content type and target mode, clearly indicating options like diagram ASCII, diagram source, or math image.
BUGFIX
Text No Longer Truncated in Narrow Terminal Windows with Auto-Wrapping for Long Lines and Tabs
When shrinking a terminal window down to a narrow width of around 40 columns, text in various panels is no longer clipped by boundary overflows. Previously, in workspace trust dialogs, the /help panel, and login or MCP authentication screens, long text frequently disappeared off the edge of the screen. Now, long lines and guidance prompts across these screens wrap automatically, and the tab bar in /help compresses its layout to fit completely, fixing text truncation issues.
原始 CHANGELOG
Original changelog遇到模型限流聰明等待，額度耗盡立刻停手
Improved rate-limit handling when the model API returns a retry delay: the CLI now waits the delay the server asks for instead of a fixed 5 seconds, and stops right away instead of retrying when the delay is longer than 30 seconds or the quota is a daily or billing capOriginal changelog提升對話畫面繪製效能，捲動與運作時更省資源
Improved rendering efficiency, cutting CPU use and memory churn while the conversation view redraws, such as during fast scrolling or while the agent is workingOriginal changelog產物檢視器按鍵提示更明確，切換模式一目了然
Improved the artifact viewer's `m` hint to name the content and the mode it switches to, such as `diagram ASCII`, `diagram source`, or `math image`, instead of `toggle ASCII` or `toggle raw`Original changelog窄終端機視窗文字不再被截斷，長行與標籤自動排版
Fixed the workspace trust dialog, the `/help` panel, and the sign-in and MCP authentication screens clipping text on narrow terminals of around 40 columns; long lines and navigation hints now wrap, and the `/help` tab bar compacts to fit